Skip to content

[Bioc-devel] Warnings in build for BioCor

8 messages · Lluís Revilla, Matthias Döring, Robert Castelo +1 more

#
Dear bioconductor core,

The development version of my package hosted in Github cannot be build. It
seems that another package is loaded when it shouldn't and masks the
functions that are used in the vignette. (In case anyone wants to help here
is the link to the StackOverflow question:
https://stackoverflow.com/q/49002455/2886003)

Additionally the current version hosted in Bioconductor has a warning in
the windows build. Which I hoped to correct for these release.

I don't know if I will find a solution on time for the next release (which
I expect in three weeks), but I am worried I won't be able to fix it on
time.

The End Of Life page of Bioconductor only talks about errors not warnings,
and I couldn't find if this has been discussed previously in the archive. I
would like to know if I can leave the development version as is (until I
find a solution) or should I aim to correct the warning for the next
release?

Many thanks,

Llu?s Revilla
#
DearLluis,
how is your namespace defined? Do you use 'import' or 'importFrom' to 
load the package in question? If you are currently using 'import' to 
load all exports from the package, you could switch to 'importFrom' and 
exclude the function that causes the problems and then use the 
double-colon operator when referring to the package's function.

Kind regards
 ?Matthias
On 02/28/2018 11:31 AM, Llu?s Revilla wrote:

  
    
#
Lluis,

something you can do in you NAMESPACE to follow the advice below from 
Matthias if you are importing the functionality of a whole package is:

import(pkgyouneed, except=functionmethodyoudefine)

in this way, you're importing everything except that particular function 
or method that you are defining in your own package.

cheers,

robert.
On 28/02/2018 12:51, Matthias D?ring wrote:
#
Dear Matthias and Robert,

My namespace doesn't import that package. It is mentioned in the
Description file in Suggests, and on the same vignette it is and loaded but
after the line where the error is reported.
The error is reported on lines 143-148 (BioCor.Rmd) and the library is
loaded on line 260.
But I might end up using the double colon to avoid the error.
Cheers,

Llu?s
On 28 February 2018 at 14:26, Robert Castelo <robert.castelo at upf.edu> wrote:

            

  
  
#
On 02/28/2018 05:31 AM, Llu?s Revilla wrote:
This means that 'another package' is on the search() path. It only gets 
there if your package attaches it, directly or indirectly. Maybe another 
package that you attach previously import'ed it, but now depends on it? 
To narrow this down, step through your vignette code until you see the 
offending package on the search() path.

Maybe you can avoid attaching the intermediate package, or put your 
package in front of the others on the search path by loading your 
package last in the vignette.

If you're running into these problems, and you're intimately familiar 
with your package, then one can imagine that users will also run into 
this problem and the suggestion to mangle your package function names 
starts to sound appealing.
The next Bioconductor release has not been announced (we try to tie 
these to the R release schedule, and the release of R-3.5 has not yet 
been announced), but will likely be at the end of April.

Martin
This email message may contain legally privileged and/or...{{dropped:2}}
#
Actually, looking at your github repository, I see you have two 
vignettes, you have renamed one from vignette2.Rmd to 
BioCor_advanced.Rmd, and that this vignette now collates before the 
original one. I believe that vignettes are built in the same R process, 
so the packages loaded in the first vignette are available (and conflict 
with) packages loaded in the second.

Martin
On 02/28/2018 08:53 AM, Martin Morgan wrote:
This email message may contain legally privileged and/or...{{dropped:2}}
#
Thanks Martin for looking into it.
I think that is the reason of the problem, the timeline of the errors
building certainly match.
I couldn't find this documented in Writing R Extensions, is this behavior
documented somewhere?

Llu?s

On 28 February 2018 at 15:03, Martin Morgan <martin.morgan at roswellpark.org>
wrote:

  
  
#
On 02/28/2018 09:35 AM, Llu?s Revilla wrote:
I'm not sure if its documented anywhere; the code here

 
https://github.com/wch/r-source/blob/trunk/src/library/tools/R/build.R#L315-L324

seems to involve processing all vignettes with a single call to create a 
new process.

Martin
This email message may contain legally privileged and/or...{{dropped:2}}